{
  "best_metric": 0.5072147846221924,
  "best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-1470_batch_1_2024-04-20_ppid_9/checkpoint-90",
  "epoch": 0.3673469387755102,
  "eval_steps": 30,
  "global_step": 90,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.04,
      "grad_norm": 83.44550323486328,
      "learning_rate": 6.666666666666667e-07,
      "logits/chosen": -1.7825796604156494,
      "logits/rejected": -1.7886196374893188,
      "logps/chosen": -169.71481323242188,
      "logps/rejected": -167.6764373779297,
      "loss": 0.6937,
      "rewards/accuracies": 0.20000000298023224,
      "rewards/chosen": -0.0002815247280523181,
      "rewards/margins": -0.0010146332206204534,
      "rewards/rejected": 0.0007331084925681353,
      "step": 10
    },
    {
      "epoch": 0.08,
      "grad_norm": 87.56982421875,
      "learning_rate": 1.777777777777778e-06,
      "logits/chosen": -1.623265027999878,
      "logits/rejected": -1.6968040466308594,
      "logps/chosen": -128.7312469482422,
      "logps/rejected": -184.20187377929688,
      "loss": 0.6912,
      "rewards/accuracies": 0.5,
      "rewards/chosen": 0.0030528828501701355,
      "rewards/margins": 0.004004458896815777,
      "rewards/rejected": -0.0009515761630609632,
      "step": 20
    },
    {
      "epoch": 0.12,
      "grad_norm": 113.4741439819336,
      "learning_rate": 2.888888888888889e-06,
      "logits/chosen": -1.7152118682861328,
      "logits/rejected": -1.758681058883667,
      "logps/chosen": -146.91494750976562,
      "logps/rejected": -213.5966339111328,
      "loss": 0.6781,
      "rewards/accuracies": 1.0,
      "rewards/chosen": 0.03251449763774872,
      "rewards/margins": 0.030697176232933998,
      "rewards/rejected": 0.0018173219868913293,
      "step": 30
    },
    {
      "epoch": 0.12,
      "eval_logits/chosen": -1.795080304145813,
      "eval_logits/rejected": -1.8307874202728271,
      "eval_logps/chosen": -194.3749542236328,
      "eval_logps/rejected": -243.1332244873047,
      "eval_loss": 0.6762310266494751,
      "eval_rewards/accuracies": 0.75,
      "eval_rewards/chosen": 0.05039369314908981,
      "eval_rewards/margins": 0.03467999026179314,
      "eval_rewards/rejected": 0.01571369729936123,
      "eval_runtime": 26.2365,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 30
    },
    {
      "epoch": 0.16,
      "grad_norm": 50.695308685302734,
      "learning_rate": 4.000000000000001e-06,
      "logits/chosen": -1.9357620477676392,
      "logits/rejected": -1.9375683069229126,
      "logps/chosen": -183.8218231201172,
      "logps/rejected": -115.80632019042969,
      "loss": 0.6768,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 0.056698836386203766,
      "rewards/margins": 0.03413978964090347,
      "rewards/rejected": 0.02255905233323574,
      "step": 40
    },
    {
      "epoch": 0.2,
      "grad_norm": 71.6834716796875,
      "learning_rate": 4.996491228070176e-06,
      "logits/chosen": -1.7773185968399048,
      "logits/rejected": -1.8331743478775024,
      "logps/chosen": -167.78533935546875,
      "logps/rejected": -227.0582733154297,
      "loss": 0.6205,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.1449195146560669,
      "rewards/margins": 0.15578031539916992,
      "rewards/rejected": -0.010860783979296684,
      "step": 50
    },
    {
      "epoch": 0.24,
      "grad_norm": 93.83502197265625,
      "learning_rate": 4.96140350877193e-06,
      "logits/chosen": -1.7739536762237549,
      "logits/rejected": -1.8414790630340576,
      "logps/chosen": -153.57481384277344,
      "logps/rejected": -194.3485870361328,
      "loss": 0.5918,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.2464807778596878,
      "rewards/margins": 0.22958576679229736,
      "rewards/rejected": 0.016895027831196785,
      "step": 60
    },
    {
      "epoch": 0.24,
      "eval_logits/chosen": -1.7974774837493896,
      "eval_logits/rejected": -1.8332849740982056,
      "eval_logps/chosen": -193.71743774414062,
      "eval_logps/rejected": -243.05783081054688,
      "eval_loss": 0.5998325943946838,
      "eval_rewards/accuracies": 0.7857142686843872,
      "eval_rewards/chosen": 0.24764502048492432,
      "eval_rewards/margins": 0.20930658280849457,
      "eval_rewards/rejected": 0.03833846375346184,
      "eval_runtime": 26.2349,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 60
    },
    {
      "epoch": 0.29,
      "grad_norm": 109.71602630615234,
      "learning_rate": 4.926315789473685e-06,
      "logits/chosen": -1.8950660228729248,
      "logits/rejected": -1.9148075580596924,
      "logps/chosen": -169.89028930664062,
      "logps/rejected": -177.87301635742188,
      "loss": 0.5992,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": 0.15271134674549103,
      "rewards/margins": 0.21031668782234192,
      "rewards/rejected": -0.0576053187251091,
      "step": 70
    },
    {
      "epoch": 0.33,
      "grad_norm": 90.3881607055664,
      "learning_rate": 4.8912280701754386e-06,
      "logits/chosen": -1.798600196838379,
      "logits/rejected": -1.855521559715271,
      "logps/chosen": -160.3217315673828,
      "logps/rejected": -247.6388397216797,
      "loss": 0.5551,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 0.29141485691070557,
      "rewards/margins": 0.3369868993759155,
      "rewards/rejected": -0.04557197913527489,
      "step": 80
    },
    {
      "epoch": 0.37,
      "grad_norm": 147.27755737304688,
      "learning_rate": 4.856140350877193e-06,
      "logits/chosen": -1.8193047046661377,
      "logits/rejected": -1.8773219585418701,
      "logps/chosen": -155.37123107910156,
      "logps/rejected": -245.6790008544922,
      "loss": 0.4932,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": 0.5351110696792603,
      "rewards/margins": 0.5927553176879883,
      "rewards/rejected": -0.05764435604214668,
      "step": 90
    },
    {
      "epoch": 0.37,
      "eval_logits/chosen": -1.8004387617111206,
      "eval_logits/rejected": -1.836446762084961,
      "eval_logps/chosen": -192.66908264160156,
      "eval_logps/rejected": -242.95899963378906,
      "eval_loss": 0.5072147846221924,
      "eval_rewards/accuracies": 0.8214285969734192,
      "eval_rewards/chosen": 0.5621528029441833,
      "eval_rewards/margins": 0.4941706657409668,
      "eval_rewards/rejected": 0.06798211485147476,
      "eval_runtime": 26.2422,
      "eval_samples_per_second": 1.067,
      "eval_steps_per_second": 1.067,
      "step": 90
    }
  ],
  "logging_steps": 10,
  "max_steps": 1470,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 90,
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}